Re: [PATCH v4] tpm: fix Atmel TPM crash caused by too frequent queries

[Hao Wu <hao.wu@rubrik.com>]

> On Aug 26, 2021, at 5:35 PM, Hao Wu <hao.wu@rubrik.com> wrote:
> 
>> On Aug 26, 2021, at 9:24 AM, Jarkko Sakkinen <jarkko@kernel.org> wrote:
>> 
>> On Wed, 2021-08-25 at 22:38 -0700, Hao Wu wrote:
>>>> On Aug 14, 2021, at 3:25 PM, Hao Wu <hao.wu@rubrik.com> wrote:
>>>> 
>>>> The Atmel TPM 1.2 chips crash with error
>>>> `tpm_try_transmit: send(): error -62` since kernel 4.14.
>>>> It is observed from the kernel log after running `tpm_sealdata -z`.
>>>> The error thrown from the command is as follows
>>>> ```
>>>> $ tpm_sealdata -z
>>>> Tspi_Key_LoadKey failed: 0x00001087 - layer=tddl,
>>>> code=0087 (135), I/O error
>>>> ```
>>>> 
>>>> The issue was reproduced with the following Atmel TPM chip:
>>>> ```
>>>> $ tpm_version
>>>> T0  TPM 1.2 Version Info:
>>>> Chip Version:        1.2.66.1
>>>> Spec Level:          2
>>>> Errata Revision:     3
>>>> TPM Vendor ID:       ATML
>>>> TPM Version:         01010000
>>>> Manufacturer Info:   41544d4c
>>>> ```
>>>> 
>>>> The root cause of the issue is due to the TPM calls to msleep()
>>>> were replaced with usleep_range() [1], which reduces
>>>> the actual timeout. Via experiments, it is observed that
>>>> the original msleep(5) actually sleeps for 15ms.
>>>> Because of a known timeout issue in Atmel TPM 1.2 chip,
>>>> the shorter timeout than 15ms can cause the error described above.
>>>> 
>>>> A few further changes in kernel 4.16 [2] and 4.18 [3, 4] further
>>>> reduced the timeout to less than 1ms. With experiments,
>>>> the problematic timeout in the latest kernel is the one
>>>> for `wait_for_tpm_stat`.
>>>> 
>>>> To fix it, the patch reverts the timeout of `wait_for_tpm_stat`
>>>> to 15ms for all Atmel TPM 1.2 chips, but leave it untouched
>>>> for Ateml TPM 2.0 chip, and chips from other vendors.
>>>> As explained above, the chosen 15ms timeout is
>>>> the actual timeout before this issue introduced,
>>>> thus the old value is used here.
>>>> Particularly, TPM_ATML_TIMEOUT_WAIT_STAT_MIN is set to 14700us,
>>>> TPM_ATML_TIMEOUT_WAIT_STAT_MIN is set to 15000us according to
>>>> the existing TPM_TIMEOUT_RANGE_US (300us).
>>>> The fixed has been tested in the system with the affected Atmel chip
>>>> with no issues observed after boot up.
>>>> 
>>>> References:
>>>> [1] 9f3fc7bcddcb tpm: replace msleep() with usleep_range() in TPM
>>>> 1.2/2.0 generic drivers
>>>> [2] cf151a9a44d5 tpm: reduce tpm polling delay in tpm_tis_core
>>>> [3] 59f5a6b07f64 tpm: reduce poll sleep time in tpm_transmit()
>>>> [4] 424eaf910c32 tpm: reduce polling time to usecs for even finer
>>>> granularity
>>>> 
>>>> Fixes: 9f3fc7bcddcb ("tpm: replace msleep() with usleep_range() in TPM 1.2/2.0 generic drivers")
>>>> Link: https://patchwork.kernel.org/project/linux-integrity/patch/20200926223150.109645-1-hao.wu@rubrik.com/
>>>> Signed-off-by: Hao Wu <hao.wu@rubrik.com>
>>>> ---
>>>> v4:
>>>> - Move timeout constants to drivers/char/tpm/tpm_tis_core.h
>>>> - Cleanup unnecessary inline comment
>>>> 
>>>> v3:
>>>> - removes unnecessary condition check in `wait_for_tpm_stat`
>>>> 
>>>> v2:
>>>> - follow the existing way to define two timeouts (min and max)
>>>> for ATMEL chip, thus keep the exact timeout logic for 
>>>> non-ATEML chips.
>>>> - limit the timeout increase to only ATMEL TPM 1.2 chips,
>>>> because it is not an issue for TPM 2.0 chips yet.
>>>> 
>>>> Test Plan:
>>>> - Run fixed kernel with ATMEL TPM chips and see crash
>>>> has been fixed.
>>>> - Run fixed kernel with non-ATMEL TPM chips, and confirm
>>>> the timeout has not been changed.
>>>> 
>>>> drivers/char/tpm/tpm_tis_core.c | 13 +++++++++++--
>>>> drivers/char/tpm/tpm_tis_core.h |  2 ++
>>>> include/linux/tpm.h             |  3 +++
>>>> 3 files changed, 16 insertions(+), 2 deletions(-)
>>>> 
>>>> diff --git a/drivers/char/tpm/tpm_tis_core.c b/drivers/char/tpm/tpm_tis_core.c
>>>> index 55b9d3965ae1..24605f100e96 100644
>>>> --- a/drivers/char/tpm/tpm_tis_core.c
>>>> +++ b/drivers/char/tpm/tpm_tis_core.c
>>>> @@ -80,8 +80,8 @@ static int wait_for_tpm_stat(struct tpm_chip *chip, u8 mask,
>>>> 		}
>>>> 	} else {
>>>> 		do {
>>>> -			usleep_range(TPM_TIMEOUT_USECS_MIN,
>>>> -				     TPM_TIMEOUT_USECS_MAX);
>>>> +			usleep_range(chip->timeout_wait_stat_min,
>>>> +				     chip->timeout_wait_stat_max);
>>>> 			status = chip->ops->status(chip);
>>>> 			if ((status & mask) == mask)
>>>> 				return 0;
>>>> @@ -934,6 +934,8 @@ int tpm_tis_core_init(struct device *dev, struct tpm_tis_data *priv, int irq,
>>>> 	chip->timeout_b = msecs_to_jiffies(TIS_TIMEOUT_B_MAX);
>>>> 	chip->timeout_c = msecs_to_jiffies(TIS_TIMEOUT_C_MAX);
>>>> 	chip->timeout_d = msecs_to_jiffies(TIS_TIMEOUT_D_MAX);
>>>> +	chip->timeout_wait_stat_min = TPM_TIMEOUT_USECS_MIN;
>>>> +	chip->timeout_wait_stat_max = TPM_TIMEOUT_USECS_MAX;
>>>> 	priv->phy_ops = phy_ops;
>>>> 	dev_set_drvdata(&chip->dev, priv);
>>>> 
>>>> @@ -983,6 +985,13 @@ int tpm_tis_core_init(struct device *dev, struct tpm_tis_data *priv, int irq,
>>>> 
>>>> 	priv->manufacturer_id = vendor;
>>>> 
>>>> +	if (priv->manufacturer_id == TPM_VID_ATML &&
>>>> +		!(chip->flags & TPM_CHIP_FLAG_TPM2)) {
>>>> +		/* If TPM chip is 1.2 ATMEL chip, timeout need to be relaxed*/
>>>> +		chip->timeout_wait_stat_min = TPM_ATML_TIMEOUT_WAIT_STAT_MIN;
>>>> +		chip->timeout_wait_stat_max = TPM_ATML_TIMEOUT_WAIT_STAT_MAX;
>>>> +	}
>>>> +
>>>> 	rc = tpm_tis_read8(priv, TPM_RID(0), &rid);
>>>> 	if (rc < 0)
>>>> 		goto out_err;
>>>> diff --git a/drivers/char/tpm/tpm_tis_core.h b/drivers/char/tpm/tpm_tis_core.h
>>>> index 9b2d32a59f67..2e431beb44f7 100644
>>>> --- a/drivers/char/tpm/tpm_tis_core.h
>>>> +++ b/drivers/char/tpm/tpm_tis_core.h
>>>> @@ -54,6 +54,8 @@ enum tis_defaults {
>>>> 	TIS_MEM_LEN = 0x5000,
>>>> 	TIS_SHORT_TIMEOUT = 750,	/* ms */
>>>> 	TIS_LONG_TIMEOUT = 2000,	/* 2 sec */
>>>> +	TPM_ATML_TIMEOUT_WAIT_STAT_MIN = 14700,	/* usecs */
>>>> +	TPM_ATML_TIMEOUT_WAIT_STAT_MAX = 15000,	/* usecs */
>>>> };
>> 
>> I'd prefer TIS_TIMEOUT_{MIN, MAX}_ATML. I.e. no "WAIT_STAT" and without "TPM_"
>> to be consistent with other constants here.
> Ok will do
>> 
>>>> 
>>>> /* Some timeout values are needed before it is known whether the chip is
>>>> diff --git a/include/linux/tpm.h b/include/linux/tpm.h
>>>> index aa11fe323c56..171b9102c976 100644
>>>> --- a/include/linux/tpm.h
>>>> +++ b/include/linux/tpm.h
>>>> @@ -150,6 +150,8 @@ struct tpm_chip {
>>>> 	bool timeout_adjusted;
>>>> 	unsigned long duration[TPM_NUM_DURATIONS]; /* jiffies */
>>>> 	bool duration_adjusted;
>>>> +	unsigned int timeout_wait_stat_min; /* usecs */
>>>> +	unsigned int timeout_wait_stat_max; /* usecs */
>> 
>> Please rename as timeout_{min, max}.
> Ok will do
>> 
>> And I think tpm_chip is wrong place to put them as they are TIS
>> specific, i.e. they should be in tpm_tis_data.
> Sorry, I am not familiar with tpm_tis_data, could tell the the place that you want me to put the var? 
> I think I may have hard time to move forward according toward this comment due to bandwidth constraints.
> Some helps would be appreciated. 
> 
> Is tpm_tis_data something specific to a chip instance ? Given the values are tied to chip,
> we need chip specific instance to make this work.

Hi Jarkko, I have checked about your proposal a bit. It look slike we need to 
Run “struct tpm_tis_data *priv = dev_get_drvdata(&chip->dev)” in every wait_for_tpm_stat call. Would this be a performance concern ? 
If we cache this in tpm_chip instance, it is not the case. 

Please let me know your thought.

Hao 

>> 
>>>> 
>>>> 	struct dentry *bios_dir[TPM_NUM_EVENT_LOG_FILES];
>>>> 
>>>> @@ -269,6 +271,7 @@ enum tpm2_cc_attrs {
>>>> #define TPM_VID_INTEL    0x8086
>>>> #define TPM_VID_WINBOND  0x1050
>>>> #define TPM_VID_STM      0x104A
>>>> +#define TPM_VID_ATML     0x1114
>>>> 
>>>> enum tpm_chip_flags {
>>>> 	TPM_CHIP_FLAG_TPM2		= BIT(1),
>>>> -- 
>>>> 2.29.0.vfs.0.0
>>>> 
>>> 
>>> Just kindly remind this code review in case it has been missed somehow
>> 
>> I'm sorry, my bad. I managed to somehow miss this. Might be because
>> I've been recently reorganizing my email accounts. And thanks for
>> pinging so that I spotted it.
> No worries, thanks for quick response!
> 
>> 
>>> Thanks
>>> Hao
>> 
>> /Jarkko
> 
> Hao